agentleFS
cmdr-characmdr-chara/chara-toolkit/skills/toolchain-preflight/SKILL.md

toolchain-preflight

Skill2 starsChanged 12 days agoMITView on GitHub

Resolve local shell, runtime, package-manager, browser, quoting, or encoding blockers before retrying the real task. Use when the execution path itself is unknown or broken. Not for application bugs or planned runtime migrations.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

11 rules. About 0 are common, 9 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use this skill when required executables or invocation behavior are unknown, or a task is blocked by shell/runtime/package-manager shims, …

    line 12
  2. Distinctive:

    Do not trigger when the repository's documented command already reaches the real task, for application debugging, deliberate …

    line 14
  3. Very specific:

    Prefer existing documented/system/runtime paths before installs or environment mutation.

    line 23
  4. Distinctive:

    Do not repeatedly retry an unchanged failing mechanism.

    line 24
  5. Distinctive:

    Use scratch locations for probes that should not affect the repository.

    line 25
  6. Distinctive:

    Do not bypass execution policy/security controls merely to make a command run.

    line 26
  7. Distinctive:

    Do not fix shell quoting, Windows command shims, or native-command invocation by nesting interpreters around destructive commands.

    line 27
  8. Distinctive:

    Establish the documented path: Read repository instructions and exact command first.

    line 31
  9. Distinctive:

    Select one safe invocation: Prefer the least-mutating supported path.

    line 34
  10. Distinctive:

    If every supported invocation requires an unauthorized install or policy change, stop with the exact blocker and the smallest required user …

    line 44
  11. Distinctive:

    Stop as soon as one supported invocation reliably reaches the actual task, or when a concrete environment/policy blocker prevents that …

    line 48

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Toolchain Preflight

Resolve one supported local execution path, then get out of the way.

Trigger boundary

Use this skill when required executables or invocation behavior are unknown, or a task is blocked by shell/runtime/package-manager shims, incompatible local runtimes, quoting, browser binaries/capability, or text encoding.

Do not trigger when the repository's documented command already reaches the real task, for application debugging, deliberate dependency/runtime migrations, or final release judgment.

Required inputs

Resolve repository instructions, working-tree/user state, OS/shell, documented command, required runtime/executable/capability, first exact failure, and any policy constraints on installs/downloads.

Safety baseline

  • Preserve uncommitted user work.
  • Prefer existing documented/system/runtime paths before installs or environment mutation.
  • Do not repeatedly retry an unchanged failing mechanism.
  • Use scratch locations for probes that should not affect the repository.
  • Do not bypass execution policy/security controls merely to make a command run.
  • Do not fix shell quoting, Windows command shims, or native-command invocation by nesting interpreters around destructive commands. Apply the cross-platform destructive-operation boundary in the toolkit workflow catalog first.

Workflow

  1. Establish the documented path. Read repository instructions and exact command first.
  2. Resolve real executables. Distinguish shell shims/wrappers from underlying runtimes and package managers.
  3. Classify the failure. Missing executable, blocked shim, wrong version/architecture, stale virtual environment, quoting, browser capability, encoding, or actual application failure.
  4. Select one safe invocation. Prefer the least-mutating supported path.
  5. Probe once, then run the real task. A successful version probe is not enough; confirm the invocation reaches the repository task.
  6. Hand off immediately. Once the execution path works, application behavior belongs to its specialist.

Handoffs and interaction boundaries

Application failures go to debugging-investigator; planned version transitions to codebase-evolution-controller; browser/app implementation to the owning builder; final integrated evidence to verification-and-release.

Failure handling

If every supported invocation requires an unauthorized install or policy change, stop with the exact blocker and the smallest required user action. If the real command runs and then fails inside the application, preflight is complete.

Stop conditions

Stop as soon as one supported invocation reliably reaches the actual task, or when a concrete environment/policy blocker prevents that without unauthorized changes.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.